The International Journal of Engineering and Science (IJES) || Volume || 6 || Issue || 5 || Pages || PP 08-15 || 2017 || ISSN (e): 2319 – 1813 ISSN (p): 2319 – 1805
A Computer Based Artificial Neural Network Controller with Interactive Auditory Guidance Ahmed Jabbar Abid1*, Alaa Desher Farhood1, Maham Kamil Naji2 1
Middle Technical University, Technical Instrurctors Training Institute, Baghdad, Iraq 2 Middle Technical University, Institute of Techonology , Baghdad, Iraq
----------------------------------------------------------ABSTRACT----------------------------------------------------------The proposed design offers a complete online and offline solution to manage the industrial systems. The designed hardware able to, read analog signals, digital signals, and controls many devices in real time. The heart of the hardware part is microcontroller PIC18F4550 which communicate with a computer via USB. The software part is programmed using Visual C# software to control managed system requires. The system operator can monitor system and diagnostic faults manually or automatically based on artificial neural network. Finally, the system has been simulated and implemented successfully. ----------------------------------------------------------------------------------------------------------------------------- ---------Date of Submission: 29 April 2017 Date of Accepted: 08 May 2017 ----------------------------------------------------------------------------------------------------------------------------- ----------
I. INTRODUCTION Industrial systems managed based on sensor readings which translated by the system; these signals are processed by the system, then controlled all the other actuators, motors, alarms, etc. The sensor’s signal could be in the form of digital or analog signals. Digital signal has a voltage level of 0 or 5 V, but analog signals have different voltage ranges like 0-5V, 0-10V or 4-20mA. The proposed system is designed to read eight analog signals, eight digital signals, and controls thirteen devices in real time. All the measured data are displayed on the computer screen and it’s possible to run any device automatically or manually. Automatic mode can be written by the system expert user to manage the entire readings base on the process requirements and the previous experience in the system. Manual mode is should control by the operator, the system will follow the operator orders to run any device, but it still needs a password to log in. A dummy operation has been taken into consideration in manual mode. The operator will be surveillance by the system software for any wrong command. The auditory warning voice played on the computer using a saved message in the C#. These messages can be changed easily according to the requirements. For now, the design has some auditory warnings like if any analog sensor goes below 10% or over 90%. Also, it’s notified the operator about any digital sensor status update. The designed system is automatically predicting and detecting any fault according based on soft implemented artificial neural network (ANN). Weights and biases are measured separately, then used by the designed software. The system designed to save data periodically to shared text file which have a privilege to read and controls remotely by the user.
II. DECISION-MAKING NETWORKS Human experience through his long journey at any skills could be concluded on his way to analyze its system measurements to detect any fault. So, if we can build such system that making decision based on learning data and can classify and predict any fault. The good news that this system can be built softly and the presented system adopts Single-layer Perceptron Classifiers (SPC). This neural network can be trained to predict and possibly fault according to our digital or analog inputs as a pattern classifier as in Fig. 1 then makes decision.
Fig. 1 Pattern classifier
DOI: 10.9790/1813-0605010815
www.theijes.com
Page 8